I wonder how hard it would be to make a 'sentence anonymizer'. Like an English to English version of Google LSTM translator network. Perhaps trained adversarially.
One example is linked in the bottom of the article:
> How to Protect Yourself: There is a project on Github you can join to help create a way to write without fingerprints: https://github.com/psal/anonymouth